Datasets:

File size: 1,363 Bytes
4a83b83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
---
pretty_name: "Flickr30k"
license: "unknown"
language:
  - en
tags:
  - image
  - text
  - captioning
  - retrieval
task_categories:
  - image-to-text
  - text-to-image
---

# Flickr30k

This repository hosts a copy of the widely used **Flickr30k** dataset,
a benchmark for image captioning and image–text retrieval.

Flickr30k contains images paired with multiple human-written English captions per image and is widely used in multimodal research.

## Contents
Commonly included files/fields (exact filenames may vary):

- Images (or image references)
- Captions (often 5 captions per image)
- Optional split / id fields

A typical record conceptually looks like:
```json
{
  "image": "<image>",
  "caption": ["...", "...", "...", "...", "..."]
}
```

## Usage
```python
from datasets import load_dataset
ds = load_dataset("nlphuji/flickr30k")
```

## Source
This dataset is taken from the Hugging Face dataset repository:
https://huggingface.co/datasets/nlphuji/flickr30k

## License
The Hugging Face dataset card does not clearly specify a single unified license (“unknown” is shown in some metadata).
Because the images originate from Flickr, usage is typically subject to the relevant **Flickr Terms & Conditions** and any per-image licensing.
Please review the dataset’s original homepage and usage terms before redistribution or commercial use.